Round trip time-based direct positioning of user equipment
The technology described herein is directed towards training an AI/ML (artificial intelligence/machine learning) model with round trip time data that captures various properties of a planned deployment of transmit-receive points. The model is trained on round-trip time measurements of communications between training device instances and transmit-receive points in an actual or simulated deployment environment. Once the model is trained, an unknown location of a user equipment in the deployment environment is determined by the trained model, by obtaining a vector dataset (acting as a ‘fingerprint’) of measured round trip times of communications between the user equipment and the transmit-receive points, and inputting the vector dataset to the trained model.
1 . A system, comprising:
at least one processor; and
at least one memory that stores executable instructions that, when executed by the at least one processor, facilitate performance of operations, the operations comprising:
receiving a vector dataset that comprises respective first round-trip times associated with a user equipment at an unknown location relative to a group of respective distributed transmit-receive points;
determining, based on the respective first round-trip times, an estimated location of the user equipment using a trained location management machine learning model trained with round-trip time training data representing second round-trip times of communications measured between the group of respective distributed transmit-receive points and a training device group at known locations, the round-trip time training data comprising measured non-line of sight round-trip time data and measured line of sight round-trip data based on respective communications between at least one transmit-receive point of the group of respective distributed transmit-receive points and respective devices, of the training device group, at respective known locations of the known locations, wherein the determining comprises using the round-trip training data as a fingerprint for determining coordinates of the user equipment based on processing the respective first round-trip times associated with the user equipment at the unknown location into the estimated location based on the fingerprint; and
outputting, from the trained location management machine learning model and based on the vector dataset, the estimated location of the user equipment.
2 . The system of claim 1 , wherein the transmit-receive point is a first transmit-receive point, wherein the known location is a first known location, and wherein the round-trip time training data further comprises line of sight data between a second transmit-receive point of the group of respective distributed transmit-receive points and at least one device of the respective devices, of the training device group, at a second known location of the known locations.
3 . The system of claim 1 , wherein the training device group comprises positioning reference units deployed at second known locations of the known locations.
4 . The system of claim 1 , wherein the training device group comprises at least one mobile device configured to report second known locations of the known locations via global positioning system data.
5 . The system of claim 1 , wherein the group of respective transmit-receive points and the training device group at the known locations are represented by a digital twin simulation of an environment, and wherein the round-trip time training data is based on the digital twin simulation.
6 . The system of claim 1 , wherein transmit-receive points of the group of respective transmit-receive points are spatially distributed in a deployment environment.
7 . The system of claim 6 , wherein the transmit-receive points are evenly distributed in the deployment environment.
8 . The system of claim 6 , wherein the operations further comprise refining spatial resolution of the transmit-receive points of the group of respective transmit-receive points via semi-supervised learning.
9 . The system of claim 1 , wherein the trained location management machine learning model comprises a polynomial regression model.
10 . The system of claim 1 , wherein the trained location management machine learning model comprises a deep neural network.
11 . The system of claim 1 , wherein the operations further comprise updating the trained location management machine learning model via reinforcement learning.
12 . The system of claim 1 , wherein the training device group comprises at least one mobile device moved among the second known locations of the known locations.
13 . A method, comprising:
inputting, by a system comprising at least one processor to a trained location management machine learning model, a dataset that comprises round trip time vector data measured via communications between a user equipment at an unknown location and at least some of transmit-receive points distributed at first known locations, the trained location management machine learning model having been trained via a training process using training round-trip time data between the transmit-receive points and a device group at second known locations, the round-trip time vector data comprising measured non-line of sight round-trip times corresponding to non-line of sight measurements and measured line of sight round-trip times corresponding to non-line of sight measurements;
determining, by the trained location management machine learning model of the system, a location estimation of the user equipment, wherein the determining comprises using the training round-trip time data as a fingerprint for determining coordinates of the user equipment based on processing the dataset that comprises the round trip time vector data associated with the user equipment at the unknown location into the location estimation based on the fingerprint; and
based on the determining, outputting, by the system in response to the inputting of the dataset, using the trained location management machine learning model based on the time vector dataset, the location estimation of the user equipment.
14 . The method of claim 13 , wherein the training process further comprises arranging non-line of sight transmit-receive points between a device of the device group and the non-line of sight transmit-receive points more densely than line of sight transmit-receive points between the device of the device group and the line of sight transmit-receive points.
15 . The method of claim 13 , wherein the training process further comprises moving at least one positioning reference unit among the second known locations.
16 . The method of claim 13 , wherein the training process further comprises moving at least one mobile device among the second known locations, and wherein the at least one mobile device reports each location of the second known locations.
17 . The method of claim 13 , wherein the training process further comprises obtaining labeled training data comprising labeled respective second coordinate data representing the second known locations, and respective round trip time data of respective communications between respective transmit-receive points at the first known locations and respective devices of the device group at the second known locations.
18 . A non-transitory machine-readable medium, comprising executable instructions that, when executed by at least one processor, facilitate performance of operations, the operations comprising:
obtaining a vector dataset at a trained location management machine learning model, the vector dataset comprising respective first round trip times measured based on first communications between a user equipment at an unknown location and a first group of respective transmit-receive points at respective first known locations, the trained location management machine learning model having been trained with respective coordinate data representing third known locations of respective devices of a training device group, and respective round trip time data representing second round trip times of respective training communications between a respective second group of transmit-receive points and the respective devices of the training device group at the third known locations, wherein the respective training communications comprises a non-line of sight communication and a line of sight communication;
inputting the vector dataset to the trained location management machine learning model;
determining, via the trained location management machine learning model and based on the respective first round trip time, an estimated location of the user equipment, wherein the determining comprises using the respective coordinate data as a fingerprint for determining coordinates of the user equipment based on processing the respective first round-trip times associated with the user equipment at the unknown location based on the fingerprint; and
in response to the determining, obtaining the estimated location of the user equipment from the trained location management machine learning model.
19 . The non-transitory machine-readable medium of claim 18 , wherein the respective first known locations comprise the respective second known locations.
20 . The non-transitory machine-readable medium of claim 18 , wherein the respective devices of the training device group at the third known locations comprise at least one of: a positioning reference unit, or a mobile device.